On 7/26/24 06:52, Niklas Cassel wrote: > On Thu, Jul 25, 2024 at 10:06:52PM +0530, Manivannan Sadhasivam wrote: >> >> I vary with you here. IMO EPF drivers have no business in knowing the BAR >> location as they are independent of controller (mostly except drivers like MHI). >> So an EPF driver should call a single API that just allocates/configures the >> BAR. For fixed address BAR, EPC core should be able to figure it out using the >> EPC features. >> >> For naming, we have 3 proposals as of now: >> >> 1. pci_epf_setup_bar() - This looks good, but somewhat collides with the >> existing pci_epc_set_bar() API. >> >> 2. pci_epc_alloc_set_bar() - Looks ugly, but aligns with the existing APIs. >> >> 3. pci_epc_get_bar() - Also looks good, but the usage of 'get' gives the >> impression that the BAR is fetched from somewhere, which is true for fixed >> address BAR, but not for dynamic BAR. > > pci_epc_configure_bar() ? > we could name the 'struct pci_epf_bar *' param 'conf' +1 But let's spell this out: pci_epc_configure_bar(), to be sure to avoid any possible confusion (config could mean configure or configuration...). -- Damien Le Moal Western Digital Research
